Two-sided market launch

Why does a new marketplace often have to pay one side of the market to show up at all?

A firm that paid people to accept its product would normally be a firm in trouble. Yet a new marketplace routinely does exactly that: guaranteed earnings for its first drivers, cash for its first sellers, free listings for years. And it does it while insisting the product is good. If the product is good, why does anyone need paying to try it?

Start at the moment before launch, because the whole puzzle lives there. What is a marketplace selling to a seller? Buyers. To a buyer? Sellers. So each side's willingness to join is a function of how many of the other side are already present — and at the instant of launch that number is zero on both sides.

Now ask each side what it should do. A seller looks at an empty marketplace and correctly declines. A buyer looks at the same empty marketplace and correctly declines. Notice that neither is making a mistake, and neither is uninformed. The good outcome — both sides present, both better off — exists and is reachable, and yet no individual can move toward it. That is a coordination failure, not a demand failure, and the distinction decides everything that follows.

Test the obvious remedies against it. More advertising? The problem is not that buyers do not know. Better software? A rider who believes the app is excellent still opens it and sees no cars. Both improve variables that are not the binding one. Whatever the platform does must alter the actual constraint: one side's participation must be made not contingent on the other's.

Paying is the crude, general-purpose way of doing that. If the cross-side benefit that would normally justify joining is missing, substitute cash of comparable size until it arrives. Which side gets the cash? Two properties decide. First, price sensitivity: subsidise the side whose presence is cheapest to buy. Second, cross-side value: subsidise the side whose presence does most to attract the other. The bill goes to the side that is neither — the one that will pay, and that wants the subsidised side badly. Caillaud and Jullien called the manoeuvre "divide and conquer" in 2003; the same year Rochet and Tirole gave the underlying point its sharpest form, that a platform chooses a price level and, separately, a price structure, and the split between the two sides is not neutral even when the total is unchanged.

Does paying actually work, though? Here the account has to be tested rather than admired. If cash were the mechanism, then any sufficiently funded launch would ignite, and many extremely well funded ones have not. So payment is sometimes necessary and never sufficient. Two things break it. It buys presence, and presence only converts into value if the sides can find each other — a thin national layer of both sides is worth far less than a saturated few square miles, which is why these launches are fought city by city or category by category rather than everywhere at once. And bought participants are the least attached ones: withdraw the subsidy and, unless they have meanwhile drawn in unpaid participants, they leave with it. The subsidy is only an investment if it buys a stock that outlives it.

Which is why the cheaper routes are taken wherever they exist. Give one side something worth having on its own, so its joining does not depend on the other at all — restaurant booking software sold as software, before any diner network existed. Or recruit by hand rather than by price: a few participants whose presence alone justifies joining beat a crowd bought at retail. Or seed the empty side yourself, which works and shades quickly into deception.

The most common misreading is that the free side is "the product being sold". Sometimes it is; but that is not what puts it on the free side. What puts it there is its price sensitivity and its pull on the other side, and plenty of subsidised sides are sold to nobody at all.

A second misreading worth naming as wrong: that launch pricing below cost is predatory pricing aimed at killing rivals. At launch there is usually no rival to kill — the competitor is the empty room. Predation is a real phenomenon and a different one, and reaching for it here explains the wrong thing.

This piece sits at a different point of the same machine as the collection's explanation of network effects. That one asks why value rises once a network exists; this one asks why the machine will not start, which is the same feedback loop read at zero, where it is a trap rather than an engine. Whether the launch spend is best understood as buying an asset or as buying a future monopoly is genuinely contested, and the same expenditure fits both readings — they differ only in what happens after the subsidy stops.

• Why some platforms tip to a single winner while others sustain several.
• How the same price-structure logic governs card interchange fees, and why regulators keep intervening in that split.

Every term the collection defines is gathered in the glossary.